Extra Surface Data Format Research

[yretenai]

NOT in the file header, this describes the format found in most games that specify extra data size 0xC (12) to 0x14 (16)

enum g1t_colorspace_t : uint8_t {
  G1T_COLORSPACE_LINEAR = 0,
  G1T_COLORSPACE_SRGB = 1
}

struct extra_data_t {
  uint32_t size; // 12, 16, or 20.
  float unk1; // assumption is depth, usually 0. some have it set to between 0.5 and 10.
  uint8_t unk2;
  uint8_t unk3;
  uint8_t unk4;
  g1t_colorspace_t colorspace; // bit flag?
  // if size >= 16
  int32_t ex_width; // if width is large enough to overflow bits in bit packing
  // if size >= 20
  int32_t ex_height; // if height is large enough to overflow bits in bit packing
};

the colorspace byte is used to determine if the surface is decoded in linear or sRGB colorspace.

[VitaSmith]

Here are my findings on this:

If we treat the data right after the depth as a 32-bit word in big-endian format (it always seems to be big endian), then we have:

• bit 0: Seems to be a marker for textures that are being directly mapped to a surface (for instance normal maps or overlays will not have this flag set). As far as I could see, the SRGB flag is part of the main flags from the common texture header, and not the extended data. But since this is still guesswork, I can't be completely positive on this.
• bits 16-21 / 28-31: These contain the number of textures for a texture array (DirectX 10 or later). Bits 28-31 are the low part and 16-21 are the high part. For instance a texture with 128 (= 0x80) frames (yes, I have seen one of those) would have 0x0008xxxx and a texture with 2 frames would have 0x2000xxxx. You can find a textures with multiple frames in Blue Reflection: Second Light in the pc000_scl model for instance. I am not sure if bits 22-27 are also meant to be used for texture array dimensions, or for something else.